Resistance-switching behavior is well known in the art and has been observed and studied in some metal-insulator mixtures since the mid 1970's. Reversible resistance-switching devices are currently one of the main contenders for replacing flash memory devices in future non-volatile memory applications. Such future non-volatile memory devices need to be increasingly scalable (to length scales lower than about 22 nanometer (“nm”)), at low energy operation and fabrication cost, and exhibit complementary metal-oxide-semiconductor (“CMOS”) process compatibility.
Some resistance-switching technologies may be triggered by voltage, a phenomenon called Electrical Pulse Induced Resistance (“EPIR”) switching effect. EPIR semiconductor devices are disclosed in U.S. Pat. No. 3,886,577 (Buckley). In the Buckley devices, a sufficiently high first voltage (50V) is generally applied to a semiconductor thin film in which an approximately 10 micron portion, or filament, of the film is set to a low resistivity state. The device is then typically reset to a high resistance state by the action of a second high voltage pulse. However, the number of switching cycles performed strongly affects set voltage. Thus, these devices generally exhibit high power consumption and poor cycle fatigue performance.
Other efforts in the art have investigated ferroelectric and magnetoresistive materials for non-volatile memory applications. These materials, however, tend to suffer from cycle fatigue and retention problems. Moreover, many magnetoresistive oxide devices require magnetic switching fields and require low operating temperatures.
Application of an electrical stimulus in a magnetic field to some perovskite family thin films shows useful resistive switching properties. Early efforts with perovskite materials required relatively high voltages and the EPIR effect tends to be cycle dependant. Later advances in these materials were able to create two terminal devices with two stable states with lower power consumption. However, the devices made from perovskite materials are largely incompatible with the semiconductor industry due to their crystal structures and the difficulties in manufacturing these materials on silicon substrates.
Metal oxides and other perovskite like materials have also been proposed for resistive switching memory devices. These metal oxide devices, however, suffer from incompatibility with silicon based semiconductor industry, and may also suffer from a lack of scalability.
Further, many of the aforementioned techniques and devices have not shown scalability for future devices, such as those expected to be in the 22 nm range for the year 2016.